Handling of CellID in scv.utils.merge

[ifanirene]

Dear scvelo team,

I'm trying to perform add loom files to pre-processed adata that is consisted of three samples (annotated with -1,-2,-3 in cellID). I saw in solved issue (#36) you mentioned that the merge module would first clean up the CellID before merging. If this is the case how would the sample correspondence be retained when merging the data?

Would you recommend the following workflow?

1. split adata by sample -> adata1, adata2, adata3
2. adata1=scv.utils.merge(adata1,ldata1)
   adata2=scv.utils.merge(adata2,ldata2)
   adata3=scv.utils.merge(adata3,ldata3)
3. adata = adata1.concatenate(adata2, adata3)

Thanks,
Irene

[WeilerP]

@ifanirene, I suppose your ldataX objects are the AnnData objects you get from the loom files?

scv.utils.merge only "cleans" the observation names if there are no overlap between adata.obs_names and ldata.obs_names:

scvelo/scvelo/read_load.py

Line 165 in 1659cc8

common_obs = pd.unique(adata.obs_names.intersection(ldata.obs_names))

scvelo/scvelo/read_load.py

Lines 168 to 171 in 1659cc8

	if len(common_obs) == 0:
	clean_obs_names(adata)
	clean_obs_names(ldata)
	common_obs = adata.obs_names.intersection(ldata.obs_names)

scv.utils.clean_obs_names removes unwanted parts from the observation names and saves the removed part in adata.obs['sample_batch'].

scvelo/scvelo/read_load.py

Lines 126 to 131 in 1659cc8

	if len(prefixes[0]) > 0 and len(np.unique(prefixes)) > 1:
	adata.obs["sample_batch"] = (
	pd.Categorical(prefixes)
	if len(np.unique(prefixes)) < adata.n_obs
	else prefixes
	)

You can use your workflow. Alternatively, you can also first concatenate the ldataX and then merge them with adata. Should be equivalent, though.

[ifanirene]

@WeilerP Thank you for the explanation! This makes a lot of sense. Yes the ldata are AnnData built from loom files. The sample annotation in the merged ldata and adata was slightly different so they got removed. I know how to fix it now.

Cheers,
Irene